Economic Effects of CO
2Abatement:
Evidence for Germany
Claudia Kemfert and Heinz Welsch, University of Oldenburg, Department of Economics
Although the economic effects of CO2abatement depend substantially on the degree to which capital and labor can substitute for energy, the issue of energy-capital-labor substitution is surrounded by considerable uncertainty. In this article we use econometri-cally estimated, sectorally differentiated elasticities of substitution for Germany to shed some light on this issue. The elasticity estimates are used within a dynamic multisector CGE model to assess the economic effects of CO2emission limits for Germany. In particular, we consider the implementation of emission limits by means of a carbon tax, assuming two alternative ways of tax revenue recycling, i.e., lump-sum transfer to private households versus labor cost reduction. The results are compared with results based on “standard” substitution elasticities from the literature. Because the estimated elasticities are on average higher and closer to unity than the “standard” elasticities, we get lower tax rates and tax revenues, and a more stable revenue/GDP ratio. In the case of using the tax revenue to reduce labor costs, the smaller revenue translates into a less favorable (but still positive) effect on employment and GDP. If the revenue is transferred to private households, the sensitivity of GDP with respect to the elasticities is rather negligible, whereas its various components are affected somewhat stronger. 2000 Society for Policy Modeling. Pub-lished by Elsevier Science Inc.
1. INTRODUCTION
The economic effects of CO2 abatement depend substantially
on the degree to which capital and labor can substitute for energy. In particular, the elasticities of substitution between energy, capi-tal, and labor affect the following issues: (i) How high will be the carbon tax rate required to attain a given CO2reduction target?
(ii) How stable will the revenue from such a tax be in relation to
Address correspondence to Prof. Dr. H. Welsch, University of Oldenburg, Department of Economics, 26111 Oldenburg, Germany.
Received January 1997; final draft accepted February 1998.
Journal of Policy Modeling22(6):641–660 (2000)
GDP? (iii) Will the tax revenue be sufficient to boost employment if used to reduce labor costs? (iv) How strong will be the macrosec-toral effects of CO2 abatement?
Compared to its importance, the issue of energy-capital-labor substitution is still surrounded by considerable uncertainty. In empirical energy-economy modeling this uncertainty involves two dimensions. First, when using multitier “nested” production func-tions there is disagreement in the literature as to the appropriate way of “nesting” energy, capital, and labor. Second, the numerical values of the substitution elasticities are controversial.
With respect to the nesting structure, Manne and Richels (1992) treat capital and labor as a composite input that trades off against energy, whereas Burniaux et al. (1992) argue that energy utiliza-tion requires appropriate equipment, implying that energy and capital are quasi-complements, trading off against labor. The usual practice in model building is to select the nesting structure ad hoc, and to pick the substitution elasticities from the literature, frequently without any sectoral differentiation.
In this paper we use econometrically estimated, sectorally differ-entiated elasticities of substitution for Germany to shed some light on the issue of energy-capital-labor substitution and its implica-tions for CO2abatement. We first use estimates based on
alterna-tive nesting structures to select the most appropriate structure. Then we incorporate the estimated elasticities in a dynamic multisector CGE model to assess the economic effects of CO2
emission limits implemented by means of a carbon tax. In particu-lar, we address the above-mentioned issues i–iv, assuming two alternative ways of tax revenue recycling, i.e., lump-sum transfer to private households versus labor cost reduction. The results are compared with results based on “standard” substitution elasticities from the literature that are uniform across sectors. Because the estimated elasticities are, on average, higher and closer to unity than the “standard” elasticities, we get lower tax rates and tax revenues, and a more stable revenue/GDP ratio. In the case of using the tax revenue to reduce labor costs, the smaller revenue translates into a less favorable (but still positive) effect on employ-ment and GDP. If the revenue is transferred to private households, the sensitivity of GDP with respect to the elasticities is rather negligible, whereas its various components are affected somewhat stronger. The effect of CO2 abatement on the sectoral outputs
elasticities do not necessarily imply that the economic effects of a given degree of CO2abatement become more favorable.
The next section addresses the estimation of elasticities of sub-stitution between energy, capital, and labor for Germany and the implications for the nesting structure. Section 3 provides a brief description of the CGE model used and the assumptions made, with an emphasis of the mechanisms by which the substitution elasticities influence the model results. Section 4 discusses the nonabatement baselines and the carbon tax rates and tax revenues implied by a given abatement target, whereas Section 5 addresses the macrosectoral results. Section 6 concludes.
2. SUBSTITUTION ELASTICITIES
2A. Previous Literature
Currently, most energy-economy models of the CGE variety use nested CES production functions to describe the production technology. Mostly, these production functions include the input factors energy, capital, and labor. The elasticities of substitution among these input factors are usually given exogenously.
Within such a framework, two questions arise. The first question is, which nesting structure should be used. In the literature, there is considerable variety on this issue. For instance, the GREEN model (Burniaux et al., 1992) uses a composite of energy and capital, trading off against labor, whereas the Global 2100 model (Manne and Richels, 1992) uses a composite of capital and labor, trading off against energy. The choice of a particular nesting struc-ture is usually settled by a priori reasoning. In this paper, by contrast, econometric evidence for the German industry is used to shed some light on this issue.
the other hand, some estimated elasticities between these factors are higher than one or even two (see Hazilla and Kopp, 1986; Chang, 1994).
To get an answer to both of these questions, three differently nested CES production functions are estimated for the entire German industry and for selected industrial sectors.
2B. Data
To estimate the substitution elasticities in the German industry, two different data sets are considered: (1) aggregate time series data for the entire German industry for the period 1970–88, and (2) disaggregated time series data for the period 1970–88 for seven industrial sectors: chemical industry, stone and earth, iron, nonfer-rous metal, vehicles, food, and paper. The data are taken from the German statistical office: output is taken to be the gross value added, and capital is given by the gross stock of fixed assets, both at 1991 prices. Labor input is measured by the persons employed in the industrial sector and energy input by final energy consump-tion (in coal equivalents) of the industrial sector.
2C. Estimated Equations
Three versions of a nested CES production function are esti-mated, using the econometric computer program SHAZAM (White, Wong, and Whistler, 1993):
(1) A two-level CES function with an energy/capital composite and labor, including a neutral technical progress factorm1:
Y15em1tA
3
a1(b1K2a11(12b1)E2a1) b1
a
1
1(12a1)L2b1
4
21 b
1
(2) A two-level CES function with a capital/labor composite and energy, including a neutral technical progress factorm2:
Y25em2tA
3
a2(b2K2a21(12b2)L2a2)b
2
a
2
1(12a2)E2b2
4
21 b
Table 1: Estimation Results for Entire Industry
Coefficient Standard error t-Ratio
(K/E)/L a1 0.1535 0.0337 4.554
b1 0.7271 0.7206 1.009
m1 0.0249 0.0014 17.28
(K/L)/E a2 0.2608 0.0889 2.9342
b2 0.4325 0.1252 3.4553
m2 0.0171 0.0022 7.854
(E/L)/K a3 4.9015 0.6256 7.9305
b3 0.1117 0.0533 2.0964
m3 0.0196 0.0052 3.7692
(3) A two-level CES function with an energy/labor composite and capital, including a neutral technical progress factorm3:
Y35em3tA[a3(b3E2a31(12b3)L2a3)
b3 a3
1(12a3)K2b3]
21 b3
2D. Estimation Results
Estimation for the entire German industry: The estimation for the period 1970–88 for the German industry1 as a whole yields
the results shown in Table 1.
Judging from thet-ratios, all coefficients except possiblyb1are
significant.
Table 2 shows theR2 and the Durbin-Watson statistic for the
various approaches. According to R2, the degree of explanation
is highest in the first approach. The DW values show no indication of positive or negative first-order autocorrelation of the residuals.
Table 2: Durbin-Watson (d) andR2Values
R2 DW—Statistic (d)
First approach: (K/E)/L 0.6769 1.11552 Second approach: (K/L)/E 0.6205 1.01739 Third approach: (E/L)/K 0.6680 1.33250
Table 3: Substitution Elasticities for Entire Industry
sa1 sb1
First approach: (K/E)/L 0.871 [K/E] 0.579 [(K,E)/L] Second approach: (K/L)/E 0.793 [K/L] 0.698 [(K,L)/E] Third approach: (E/L)/K 0.167 [E/L] 0.899 [(E,L)/K]
Overall, approach (1) is our preferred model for the entire German industry.
The substitution elasticities can be recovered from the estimated coefficients as follows:
sa
i5
1 11 ai
orsb ii5
1 11 bi
.
They are shown in Table 3.
The estimated coefficients all imply positive substitution elastic-ities. Thus, energy, capital and labor are substitutes in the produc-tion funcproduc-tion of the German industry.
Estimation for industrial sectors: To estimate the substitution elasticities of the three versions of the nested CES production function, time series data from 1970–88 for the individual sectors are used.
Table 4 shows the coefficients of determination for the three versions of the production function.
All R2 values are substantially higher than their counterparts
for the entire industry. The ranking of the various models differs from sector to sector.
The problem of serial autocorrelation can be neglected in all approaches because all Durbin-Watson values are not in the criti-cal area.
Table 4: SectoralR2Measures
KE/L KL/E EL/K
Chemical industry 0.9894 0.9850 0.5127 Stone and earth 0.8809 0.9990 0.8909
Nonferrous 0.8997 0.9999 0.8854
Iron 0.8653 0.9998 0.9596
Vehicle 0.9897 0.9998 0.9994
Paper 0.9996 0.9753 0.9989
Table 5: Sectoral Substitution Elasticities
K/E KE/L K/L KL/E E/L EL/K
Chemical 0.49 0.85 0.55 0.96 0.68 0.95 Stone and earth 0.98 0.94 0.54 0.91 0.90 0.94 Nonferrous 0.23 0.91 0.20 0.77 0.97 0.32 Iron 0.17 0.92 0.55 0.98 0.88 0.71 Transport 0.31 0.87 0.17 0.88 0.52 0.83 Paper 0.91 0.62 0.52 0.96 0.20 0.89 Food 0.75 0.76 0.58 0.64 0.07 0.90 Estimation for
whole Industry 0.871 0.579 0.793 0.698 0.167 0.899
The implied elasticities of substitution are shown in Table 5. Compared to the results for the entire industry we see that the sectors “stone and earth” and “paper” show a higher elasticity of substitution between capital and energy than estimated for the whole industry. The sector nonferrous metal has very low substitu-tion elasticities between capital and the other input factors. In all sectors the substitution elasticities (KE/L) and (KL/E) are higher than the corresponding elasticities for the whole industry. The substitution elasticities between capital and labor are lower and those between energy and labor are higher than for the entire industry. All other substitution elasticities differ more or less from the aggregated substitution elasticity.
Overall, it can be concluded that a nested CES production function with a composite of capital and energy seems most appro-priate for the entire industry, whereas the evidence is mixed on the sectoral level. All substitution elasticities are in the range between 0 and 1, which means that the inputs are imperfect substi-tutes. Sectorally disaggregated estimations exhibit intersectoral differences of substitution elasticities.
3. MODEL AND ASSUMPTIONS 3A. General Description
The simulation model used in this paper is LEAN-TCM, a general equilibrium model of Germany and the rest of the Euro-pean Union.2The simulation exercises described below focus on
Figure 1. Production hierarchy.
the submodel for Germany. The model has the following sectors: hard coal, lignite, petroleum, gas, electricity, agriculture, interme-diate products, equipment goods, consumption goods, construc-tion, transportaconstruc-tion, private services, and public services. The model is recursively dynamic, with a time horizon to 2020.
With respect to the effects of CO2 taxation, the modeling of
production and of wage formation are of special interest. Produc-tion is described by a five-stage nested producProduc-tion funcProduc-tion for each sector, allowing for a flexible treatment of substitution possi-bilities. Figure 1 displays the production hierarchy. At the top level, output is linked to an aggregate of energy, capital and labor (EKL) and to the various intermediate inputs via constant input– output coefficients. The EKL aggregate is further broken down into labor and an energy-capital aggregate, which is in agreement with the results of the preceding section. Next, energy-capital is separated into capital and energy. Energy, in turn, is an aggregate of fossil energy and electricity. Finally, fossil energy is a composite of hard coal, brown coal, petroleum, and gas. Typically, the elastic-ity of substitution among fossil fuels is larger than that between fossil fuels and electricity. The latter, in turn, is larger than that between energy and capital.3Factor demand is derived from profit
maximization subject to the production structure just outlined.
Concerning the labor market, it is assumed that wage rates are uniform across sectors. The current wage in each period equals the wage of the previous period times the increase in labor produc-tivity (see, e.g., Conrad and Wang 1993), modified by the ratio of actual employment to “normal” employment (Phillips curve). Employment is the sum of labor demand across sectors.
3B. Some Key Relationships
This subsection elaborates on some model relationships that are of key importance for the influence of altered elasticities of substitution. In particular, we focus on the effect of a higher elasticity of substitution between energy/capital and labor, because the estimates in Section 2 suggest that this elasticity is higher than usually assumed (see Section 3C).4 The presentation is slightly
simplified in comparison with the full model structure. Especially, the sectoral differentiation is disregarded.
LetXdenote the output produced from energy/capitalEKand laborL according to the following production function:
X5
3
dEK(aEKEK)wheres is the elasticity of substitution betweenEKand L, dEK,
anddLare the distribution parameters, andaEKandaLthe efficiency
factors. The associated demand function forEKis
EK5ds
Thus, the elasticity of substitution translates straightforwardly into the price elasticity of demand. Observe that withs ,1 a higher efficiency factoraEKreduces demand.
We will now consider how the evolution of the efficiency factor together with the pricespXand pEKinfluence the factor intensity
EK/X. To see this, it is useful to rewrite Equation 2 in terms of growth ratesg(.). This gives
g(EK/X)5 sg(pX)2(sg(pEK)1(12 s)g(aEK)). (3)
This equation shows thatEK/Xdeclines if g(pEK) and g(aEK) are
sufficiently large, relative tog(pX). Furthermore, it shows that an
increase ins reduces the rate of decline ofEK/X, provided that
g(pEK),g(aEK). When considering the numerical assumptions in
Section 3C, it will be seen that the latter is, in fact, the case. Therefore, a higher elasticity of substitution betweenEKand L
implies a slower decline of EK/X; hence, a higher EK/X at all points in time (given the initial value).
Complementary to the slower decline ofEK/Xwe have a slower
increase of the labor intensity L/X or, equivalently, a faster in-crease of the labor productivityX/L. Because the rate of growth of the labor productivity is the main determinant of the growth of the wage rate, we can conclude that under the conditions specified above (g(pEK), g(aEK), g(pX) sufficiently small) an increase ins
implies a higher growth and, hence, level of the wage rate (given the initial value).
These conclusions hold ceteris paribus, i.e., without consider-ation of CO2abatement. If conversely, we do consider CO2
abate-ment policies, it is clear that a policy-driven reduction ofEKhas less impact on labor productivity ifsand, hence, theEK/Xratio in the nonabatement baseline is higher. Thus, under a higher elasticity CO2abatement will imply a smaller reduction of wages.
These considerations will turn out to be important when consid-ering the effects of CO2 abatement under alternative elasticity
assumptions.
3C. Assumptions and Scenarios
The “standard” assumptions on substitution elasticities are as follows: energy/capital versus labor—0.6, energy versus machin-ery—0.3, energy versus structures—0.9, electricity versus fuels— 0.6, fuel versus fuel—0.8.
These “standard” assumptions are mostly near the center of the range of estimates to be found in the literature (see Burniaux et al., 1992 for an overview). They are uniform across sectors. The elasticities of substitution between energy and “capital” re-quired in the model are sector-specific averages of the energy/ machinery and the energy/structures elasticities. Thus, the overall energy-capital elasticity becomes larger as the share of structures in the overall capital stock becomes larger.
Energy/capital Energy versus versus labor machinery Intermediate products 0.85 0.49
Equipment goods 0.90 0.20
Consumption goods 0.70 0.80
Transportation 0.87 0.31
Note that these elasticities are mostly higher than the “standard” assumptions.
As discussed in the preceding subsection, the choice of substitu-tion elasticities influences the nonabatement baseline. In addisubstitu-tion, the baseline is determined by assumptions on factor productivities and the development of the exogenous variables. The productivity factor for energy is assumed to grow at 2 percent annually, and the productivity factor for labor and capital at 1.5 percent. The exogenous driving forces of the model are the import volumes and the export prices of the rest of the world. The former are assumed to grow at 2 percent, while the (real) export prices are assumed to be constant, except for the energy carriers. For the latter, an annual price increase of 1.4 and 2 percent is assumed for petroleum and gas, respectively, while the world market price for coal is assumed to be approximately constant.5Thus, average
energy prices grow less than energy efficiency.
These assumptions imply a baseline GDP growth of about 2 percent and an annual growth in CO2emissions by 0.9 percent in
West Germany under “standard” elasticity assumptions.
The CO2abatement target is specified as involving a stabilization
of emissions in 2005 through 2020 at their level in 1990. For 2000–05 a linear adjustment of emissions towards this target value is postulated. The abatement target is implemented by means of a carbon tax whose level is determined implicitly from the emission limit. The revenue from the tax can be used in two ways: REDIST refers to a lump-sum redistribution of the tax to private house-holds. EMPLOY means that the revenue is used to subsidize wages, i.e., the “purchase price” of labor is smaller than the wage rate by the amount “tax revenue/employment.”6
Table 6: Nonabatement Baseline under Alternative Elasticity Assumptions
Emissions GDP Emissions/GDP
(million tons CO2) (billion ECU) (tons/1000 ECU)
Standard elasticities
2000 835.89 1139.57 0.73
2005 878.83 1263.31 0.70
2010 920.92 1397.69 0.66
2015 963.63 1543.48 0.62
2020 999.78 1701.69 0.59
Higher elasticities
2000 914.02 1148.82 0.80
2005 960.15 1275.07 0.75
2010 1004.80 1410.45 0.71
2015 1050.33 1556.11 0.67
2020 1087.19 1713.17 0.63
4. CO2EMISSIONS AND CARBON TAX
To understand the influence of higher elasticities, it should be recalled that they change not only the effects of abatement, but, in the first place, the nonabatement emission baseline. This effect arises because world energy prices are assumed to grow at a lower rate than the autonomous energy efficiency factor. Under these circumstances, a higher elasticity of substitution between energy-capital and labor implies a higher energy intensity (see Section 3B). As a result, the baseline emission level will be higher under higher substitution elasticities.
Table 6 shows the baseline carbon emissions under alternative elasticity assumptions, along with GDP and the emission intensity (CO2/GDP). It can be seen that the difference in emissions is quite
pronounced.7 Because of the higher energy input in the
high-elasticity case, GDP is also somewhat higher.
The fact that the nonabatement baselines develop differently under alternative sets of elasticities has important implications for the macroeconomic effects of abatement, as will be seen in the next section. In the first place, however, it raises the problem of how to define “equivalent” abatement targets under alternative sets of elasticities. Obviously, requiring thesame target emission level under both sets of elasticities would imply a much higher
degree of abatement in the high-elasticity case than in the low elasticity case, which would bias our estimates of the required tax rate as well as the economic effects of abatement.8Thus, we define
an “equivalent” abatement target as one involvingequal absolute abatement. In other words, we compute how much abatement is required in the low-elasticity case to attain the emission level of 1990; then we subtract this abatement level from the high-elasticity emission baseline to obtain high-elasticity target emissions.
Overall, the following abatement scenarios are considered: A-R: stabilization of emissions at 1990 level under standard elasticities, REDIST; B-R: equal absolute abatement under higher elasticities, REDIST; A-E: stabilization of emissions at 1990 level under stan-dard elasticities, EMPLOY; B-E: equal absolute abatement under higher elasticities, EMPLOY.
The tax rates, tax revenues, and revenue shares in GDP that are associated with these abatement scenarios are shown in Table 7. Consider first the cases in which the revenue is rebated to private households (REDIST). It can be seen that in the high elasticity case significantly lower tax rates are required than under standard elasticities, for equal absolute abatement (B-R). If the tax revenue is used to reduce labor costs (EMPLOY), we also find that higher elasticities imply lower tax rates. This is what one would expect: higher elasticities of substitution mean higher price elasticities, which implies that a given abatement target requires a smaller tax-induced price increase.
If we compare the EMPLOY cases with the corresponding REDIST cases we find that the tax rates are higher in the former than in the latter. The reason for this is that under EMPLOY we have a higher level of economic activity than under REDIST (see Section 5).
The tax revenues in the various cases are an immediate conse-quence of the corresponding tax rates: higher elasticities imply lower revenues and lower revenue/GDP ratios, and EMPLOY implies higher revenues and revenue/GDP ratios than REDIST. A question that has received much attention in the public fi-nance literature on an “ecological tax reform” concerns the stabil-ity of the revenue/GDP ratio (see Pethig, 1997). In this regard we find that in all six scenarios the revenue share in GDP increases
Table 7: Tax Rates and Tax Revenues
Tax Rate Revenue Revenue/GDP
(ECU/ton CO2) (billion ECU) (percent)
Scenario A–R
2000 3.25 2.64 0.19
2005 21.88 16.17 1.31
2010 30.03 22.20 1.64
2015 39.40 29.12 1.97
2020 48.24 35.66 2.20
Scenario B–R
2000 2.56 2.28 0.20
2005 15.01 12.36 0.98
2010 20.84 17.16 1.26
2015 27.49 22.64 1.52
2020 33.53 27.67 1.69
Scenario A–E
2000 3.52 2.86 0.25
2005 24.88 18.39 1.46
2010 34.46 25.47 1.84
2015 45.65 33.74 2.22
2020 56.27 41.59 2.49
Scenario B–E
2000 2.80 2.29 0.21
2005 16.95 13.96 1.10
2010 23.64 19.47 1.40
2015 31.39 25.84 1.69
2020 38.38 31.61 1.89
over time. However, in line with theoretical expectations, the increase is less pronounced under the higher (and closer-to-unity) elasticities.
5. MACROSECTORAL EFFECTS
We consider the macrosectoral effects separately for the two types of recycling of the tax revenue.
Table 8: Macrosectoral Effects in REDIST Case, Percentage Difference from Baseline
2000 2005 2010 2015 2020
Scenario A–R
CO2emissions 22.83 215.89 219.74 223.29 226.07 GDP 0.35 20.08 20.25 20.64 21.03 Consumption 0.28 20.47 20.87 21.35 21.78 Investment 5.26 3.88 0.53 20.99 22.06 Export 20.87 21.26 20.40 20.43 20.49 Import 0.96 20.83 22.16 22.69 22.93 Employment 0.63 20.15 20.02 20.12 20.12 Hard coal 0.00 0.00 0.00 0.00 0.00 Lignite 28.91 236.77 241.81 245.68 248.43 Petroleum 20.71 26.04 28.13 210.30 211.60 Gas 20.73 26.24 29.17 212.30 214.96 Electricity 21.38 26.05 27.82 210.13 212.28 Agriculture 20.24 20.87 21.03 21.63 22.26 Intermed. products 20.60 22.37 22.30 23.13 23.89 Equipment goods 0.71 0.20 0.40 0.20 0.09 Consumption goods 20.09 20.52 20.26 20.55 20.83 Construction 2.74 1.36 20.41 21.46 22.18 Transportation 0.13 20.55 20.62 20.98 21.28 Private services 0.34 20.30 20.64 21.12 21.58 Public services 20.26 20.73 20.91 20.99 21.08 Scenario B–R
substantially less. The latter result has to be viewed in conjunction with the strong reduction of imports, which occurs because Ger-man imports contain a substantial fraction (more than 20 percent) of fossil fuels. As a result of the strong decline in imports, the exchange rate (price of foreign currency) drops. This explains why exports decrease less than domestic demand. Employment decreases less than GDP because energy is partly substituted by labor. Sectoral outputs decline by varying degrees, depending on their energy intensity.9Only the equipment-good industry shows
a small increase, which is mainly due to the initial increase in investment.
Turning to scenario B-R, i.e., the case of equal absolute abate-ment under higher elasticities, we see that the effect on GDP is practically the same as in the A-R case. However, the GDP components are now affected differently. Especially, investment is reduced more, whereas consumption is reduced less than in the previous case. This is surprising because, as discussed in the preceding section, the tax revenue, to be rebated to private house-holds, is now lower. Thus, the smaller reduction of consumption can only be explained by referring to a component of available household income other than the tax rebate. In fact, a more de-tailed inspection of the results (not shown in the table) reveals that the wage rate and, hence, the wage income is about 4 to 5 percent higher in the B-R case than in the A-R case. This is in line with the reasoning developed in Section 3B that a given amount of CO2 reduction reduces labor productivity less if the
baseline CO2intensity is higher. What remains surprising, then,
is that this effect dominates the effect of the smaller tax revenue, leading to a smaller reduction of overall household income in B-R than in A-R, and to a smaller decline in consumption.
In contrast to consumption, investment and the sectoral outputs are now more strongly reduced than in the A-R case. This is nothing more than an implication of the mechanism just sketched: the wage reduction is now smaller than in the A-R case, and this effect dominates the effect of the lower tax rate, leading to produc-tion costs being higher in the B-R case than in the A-R case.10
9Hard coal mining is not affected by CO2abatement because the output of this industry is regulated and, therefore, determined exogenously.
These considerations suggest that the effects of higher substitu-tion elasticities are rather involved.
We next consider the effects of CO2abatement when the tax
revenue is used to reduce the wage rate facing the employers (relative to consumer wages). The results are given in Table 9.
In scenario A-E (standard elasticities) this leads to an increase in GDP, consumption, investment, exports and, especially, em-ployment. Imports decrease, as previously, because of the high share of fossil fuels in German imports. The sectoral outputs increase, except for the energy sectors and the energy-intensive basic-materials industry.
In scenario B-E, i.e., equal absolute abatement under higher elasticities, all the increases are lower than in scenario A-E. This holds especially for employment and GDP. The reason for the smaller increase in economic activity levels is simply that in the high elasticity case a smaller tax revenue is available for labor cost reduction.11
6. CONCLUSIONS
In this paper we have used econometrically estimated elasticities of substitution between energy, capital, and labor to assess the general equilibrium effects of CO2 abatement in Germany.
Em-phasis was placed on comparing these effects with results obtained under “standard” elasticity assumptions picked from the literature. Because the estimated elasticities are on average higher than the “standard” assumptions they imply lower carbon tax rates and tax revenues associated with a given target amount of CO2
reduction. Also, being closer to unity than the “standard” assump-tions, the estimated elasticities imply a revenue/GDP ratio that is more stable over time.
The influence of the higher elasticities on the macrosectoral effects of CO2abatement is found to depend substantially on the
way in which the tax revenue is recycled into the economy. If the revenue is used to reduce labor costs, the lower revenue implied by higher elasticities translates into less expansionary impulses. If the revenue is rebated to private households, the impact of higher elasticities on the macrosectoral effects of abatement is
Table 9: Macrosectoral Effects in EMPLOY Case, Percentage Difference from Baseline
2000 2005 2010 2015 2020
Scenario A–E
CO2emissions 22.83 215.89 219.74 223.29 226.07
GDP 0.64 1.97 2.39 2.61 2.70
Consumption 0.49 1.12 1.25 1.32 1.34 Investment 5.83 6.19 2.97 1.91 1.20 Export 20.70 20.17 0.95 1.25 1.43 Import 1.06 20.21 21.40 21.78 21.91 Employment 1.12 3.26 4.15 4.85 5.39 Hard coal 0.00 0.00 0.00 0.00 0.00 Lignite 29.10 237.20 242.21 245.95 248.63 Petroleum 20.63 25.68 27.71 29.87 211.13 Gas 20.66 25.68 28.76 211.90 214.61 Electricity 21.32 25.28 26.78 28.94 211.04 Agriculture 20.11 0.18 0.53 0.50 0.37 Intermed. products 20.34 20.58 20.02 20.31 20.66 Equipment goods 1.16 2.87 3.63 4.16 4.58 Consumption goods 0.17 1.24 2.05 2.41 2.63 Construction 3.16 3.49 2.11 1.59 1.28 Transportation 0.37 1.10 1.53 1.71 1.81 Private services 0.58 1.44 1.63 1.71 1.69 Public services 20.02 2.01 2.77 3.48 3.93 Scenario B–E
CO2emissions 22.59 214.23 218.04 221.60 224.25
GDP 0.59 1.51 1.79 1.92 1.93
ambiguous. It depends considerably on the nonabatement base-line. Under plausible assumptions on world energy prices and autonomous energy efficiency improvements, higher substitution elasticities imply higher baseline emissions. Therefore, reducing emissions (i.e., fossil energy input) by a fixed amount reduces labor productivity substantially less under high elasticities than under low ones. Thus, wages remain higher in the high elasticity case, whereas the transfer income from the carbon tax is lower. In our simulations, the former effect dominates the latter, implying a smaller decrease of consumption under high elasticities than under low ones. On the other hand, investment and exports show a stronger decrease under high elasticities because prices are higher in this case.
It should be noted that in these considerations the economic effects of CO2 abatement are measured against nonabatement
baselines that are differentiated according to the elasticity assump-tions made. In contrast to this approach, one could argue that higher elasticities are always economically favorable, because a
given fixed amount of abatementleads to a higher economic activity level if applied to a higher rather than a lower baseline. Against such reasoning it may be objected, however, that applying a given fixed amount of target abatement equally to high and low baseline emissions is not ecologically sensible. From a strictly ecological point of view, target emissions rather than target abatement levels should be kept fixed in comparing the economic effects of abate-ment under alternative elasticity assumptions. Had we used this normalization, the effects of abatement under higher elasticities would have been less favorable than described in this paper.
REFERENCES
Berndt, E.R., and Wood, D.O. (1975) Technology, Prices and the Derived Demand for Energy.Review of Economics and Statistics56:259–268.
Burniaux, J.M., Martin, J.P., Nicoletti, G., and Oliveira-Martins, J. (1992) GREEN—A Multi-Sector, Multi-Region General Equilibrium Model for Quantifying the Costs of Curbing CO2Emissions; A Technical Manual. OECD Working Paper No. 116, Paris.
Chang, K.-P. (1994) Capital-Energy Substitution and the Multi-level CES Production Function.Energy Economics10:22–26.
Conrad, K., and Wang, J. (1993) Quantitative Umweltpolitik: Gesamtwirtschaftliche Aus-wirkungen einer CO2-Besteuerung in Deutschland (West).Jahrbu¨cher fu¨r Nation-alo¨konomie und Statistik212:309–324.
Hudson, E.A., and Jorgenson, D.W. (1974) US Energy Policy and Economic Growth, 1975–2000.The Bell Journal of Economics and Management Science5:461–514. Hunt, L.C. (1986) Energy and Capital: Substitutes or Complements? A Note on the
Importance of Testing for Non-neutral Technical Progress. Applied Economics
18:729–735.
IEA (1994)World Energy Outlook. Paris: IEA.
Kemfert, C. (1998) Estimated Substitution Elasticities of a Nested CES Production Func-tion Approach for Germany.Energy Economics20:249–264.
Magnus, J.R. (1979) Substitution between Energy and Non-Energy Inputs in the Nether-lands 1950–1976.International Economic Review20:465–484.
Manne, A.S., and Richels, R.G. (1992)Buying Greenhouse Insurance—The Economic Costs of CO2Emission Limits. Cambridge: MIT Press.
Pethig, R. (1997) Emission Tax Revenues in a Growing Economy, Discussion Paper No. 61-97, University of Siegen.
Prywes, M. (1986) A Nested CES Approach to Capital-Energy Substitution.Energy Eco-nomics1:22–28.
Welsch, H., and Hoster, F. (1995) A General Equilibrium Analysis of European Carbon/ Energy Taxation: Model Structure and Macroeconomic Results. Zeitschrift fu¨r Wirtschafts- und Sozialwissenschaften115:275–302.
